Author: Christophe Theron
Date: 22:51:12 04/02/98
Go up one level in this thread
On April 02, 1998 at 16:16:53, Amir Ban wrote:
>On April 02, 1998 at 12:31:17, Christophe Theron wrote:
>
>>Version 3.1 introduced some improvements in the avalaible optimizations.
>>
>>But one bad thing was that turning all optimizations on generated
>>incorrect code! I had, for example, to turn some optimizations off to
>>successfully compile Chess Tiger 9. This was true for both DOS and
>>Windows compilers (BC3.1 provided 2 environments). I used BC3.1 for 5
>>years.
>>
>>All in all, BC3.1 was able to produce average speed code.
>>
>
>It's a superb compiler. It dates from 1992 and was to my mind the best
>compiler Borland ever produced. In later versions they started
>neglecting 16-bit code, so I never found a reason to upgrade. I've been
>using it for the Junior DOS version for a long time. It's also still the
>standard 16-bit compiler at my place of work, and I see it being used in
>other companies including Intel.
I would have loved to see a good successor for BC3.1. But Borland didn't
provide the DOS IDE any more in the next releases, and their compilers
began to be more and more buggy.
It's really hurting me, but I have to say it: the latest BC compiler
(5.0) was unable to compile and run my program. BC4.5 could do it, so
could MSVC 4 and 5. But not BC5, even with the latest patches I
downloaded from Borland. A real shame.
Fortunately, we have the free GnuC compiler, which has exactly the same
look and feel than the BC3.1 DOS IDE. If you love BC3.1 DOS, Amir, you
might well fall in love with GnuC. And it produces fast and reliable 32
bits DOS applications. Tiger compiled with GnuC is only 5% slower than
compiled with MSVC5. And I can execute the GnuC version under pure DOS,
so the speed difference is very small (MSVC executable has to be run
under Windows).
>>But I'm very surprised that Amir used this quite old compiler to release
>>the commercial Junior code. I suppose he was forced to do so because he
>>had to run inside the Fritz 16 bits environment?
>>
>
>This was the first time I've done anything on Win16, and there was the
>consensus that Microsoft compilers are faster, so I initially developed
>on MSVC 1.51. When it was done, I found the performance disappointing.
>After many attempts to improve, without great results, I decided to give
>my old BC3.1 a try in its Win16 mode which I've never used before. It
>was equal to the Microsoft compiler on P5 but 10% faster on P2/P6.
In my experience, Microsoft compilers before version 5 did not produce
very fast code. Only MSVC5 is able to beat BC4.5 and GnuC.
Are you still using BC3.1 DOS now, Amir? For Junior or anything else?
Christophe
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.